Multicenter ex vivo evaluation of an α-mannosidase–targeted fluorescent probe for intraoperative margin assessment in breast cancer

Abstract

Determining an accurate intraoperative margin is essential in breast-conserving surgery (BCS) for breast cancer. This multicenter study assessed the diagnostic potential of a novel fluorescent probe targeting α-mannosidase, named the α-Man probe, in evaluating intraoperative margins. We analyzed 274 resected breast tissue samples, including malignant and benign lesions, across four institutions. Fluorescence intensity (FI) was measured 5 and 20 min following the application of the α-Man probe. Receiver operating characteristic analysis defined definitive thresholds for changes in FI (ΔFI), maintaining false positive and false negative rates at ≤ 3%. Diagnostic performance was also compared with a γ-glutamyltranspeptidase (GGT)-based fluorescent probe that we previously developed. Using 5-min ΔFI thresholds, positive, uncertain, and negative classifications were 24.1, 53.6, and 22.3%, respectively. At 20 min, these classifications improved to 28.1, 38.3, and 33.6%, respectively. At 5 and 20 min, positive predictive values were 93.9% and 94.8%, whereas negative predictive values were 95.1% and 96.7%, respectively. Compared with the GGT probe, α-Man substantially reduced the uncertain classification rate while maintaining high predictive accuracy. This study demonstrated that a two-step ΔFI-based protocol using the α-Man probe enables the accurate, real-time, and pathologist-independent assessment of resected specimens, supporting its future application to intraoperative margin evaluation during BCS.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-42316-2.

Introduction

Breast cancer (BC) is the most common malignancy among women, with incidence increasing annually¹. Standard treatment for BC involves breast-conserving surgery (BCS) followed by radiation therapy². An accurate intraoperative assessment of surgical margins during BCS is crucial to minimize reoperation and local recurrence rates^3–7, and intraoperative frozen section analysis (IFSA) is currently widely employed owing to its reliability^8–11. However, this technique is expensive and has certain limitations, including the need for a skilled pathologist, dedicated space, and limited throughput, which often results in false negatives^12–15. Therefore, alternative methods that enable rapid, efficient, and accurate margin assessments are needed.

Fluorescence-guided detection represents a promising alternative to IFSA¹⁶. It possesses various advantages; for example it does not require an on-site pathologist, is comparatively cheaper, easier to use, has minimal space requirements, and enables rapid analysis. Furthermore, it allows for subsequent pathological examination using the same specimen. We previously developed a γ-glutamyltranspeptidase (GGT)-reactive fluorescent probe (gGlu-HMRG) and demonstrated its use in intraoperative margin assessments^17–20. However, the sensitivity and specificity of the GGT probe were found to be insufficient, owing to the significant background fluorescence of normal breast tissue and benign proliferative lesions, which resulted in low positive detection rates and frequent inconclusive results.

To overcome these limitations, we focused on using glycosidases, which are enzymes implicated in tumor initiation, progression, and metastasis, to detect BC^21,22. We synthesized a series of fluorescent probes by attaching glycosides to our established fluorophore scaffold HMRef^23–25. Among these, the α-Man probe targeting α-mannosidase 2C1 (MAN2C1) exhibited superior specificity and sensitivity for BC visualization (Supplementary Fig. S1)²⁶. MAN2C1 is a cytoplasmic enzyme that cleaves α-1,2, α-1,3, and α-1,6-linked mannose residues from glycoproteins during free oligosaccharide degradation²⁷. Notably, MAN2C1 is overexpressed in BC cells²⁶.

In the present multicenter study, we aimed to evaluate the diagnostic potential of applying the α-Man probe to resected breast specimens, with the ultimate goal of developing its application in intraoperative surgical margin assessment during breast-conserving surgery.

Results

First exploratory study at the Ueo breast cancer hospital

The primary objective of this initial exploratory study was to evaluate the diagnostic capability of the α-Man probe for BC detection and compare it with our previously developed GGT probe^19,20. The second objective was to determine the optimal measurement time with the highest diagnostic accuracy using the α-Man probe. We evaluated 95 samples obtained from 46 participants as the full analysis set (FAS; Fig. 1). After excluding two samples with a diagnosis of mucinous carcinoma, 93 samples were prepared as the per protocol set (PPS).

Fig. 1.

Sample preparation.

Quantitative differences in 5-min ΔFI among different lesion types

In the first exploratory study, 93 samples, including 34 malignant and 59 non-malignant samples, were analyzed. The malignant lesions exhibited a time-dependent increase in fluorescence (ΔFI) that clearly distinguished them from the non-malignant tissues, which included mammary glands and fatty tissue (Fig. 2). At 5 min, the ΔFI of both IC and NIC was significantly larger (p < 0.0001, Wilcoxon rank-sum test) than that of non-malignant lesions (Fig. 3A).

Fig. 2.

Representative examples of fluorescence intensity changes. (A) ① Visible light images of the initial FI, ② 15 min FI, and ③15 min FI images with pseudo-color. ④ HE-stained images of DCIS. The black line indicates non-invasive cancer. The scale bars indicate 5 mm. (B) Fluorescence intensity changes in the areas indicated by squares in the visible light images. The surrounding lesion includes non-invasive cancer.

Fig. 3.

Quantitative differences in ΔFI among different lesion types and definitive thresholds (exploratory study). The value of ΔFI in various lesions is plotted against the pathological classification of each sample. Each sample was classified as invasive cancer (IC), non-invasive cancer (NIC), or benign lesion (Non-malignant).The solid and dashed lines represent the definitive negative and positive thresholds, respectively, as determined by data points with a 3% false negative rate and a 3% false positive rate. (A) 5 min ΔFI and (B) 20 min ΔFI.

Setting definitive thresholds

In this study, we determined the definitive thresholds using data points with a 3% false negative rate and a 3% false positive rate in the receiver operating characteristic (ROC) curve analysis, as in our previous studies^19,20. These thresholds were independently determined for each analysis group. The diagnostic range was classified into the following three groups: negative (-), below the definitive negative threshold; positive (+), exceeding the definitive positive threshold; and the uncertain group, falling between the two definitive thresholds (Fig. 3A and B).

Analysis according to the definitive threshold of 5 min ΔFI

According to the definitive negative threshold of 5 min ΔFI, 93 samples were divided into the following categories: 27 (29.0%) were positive (+), 48 (51.6%) were uncertain, and 18 (19.4%) were negative (-). Among the 34 cancer samples, 26 (76.5%) were categorized as positive (+), 7 (20.6%) as uncertain, and only 1 case (2.9%) was negative (-), and this was a small-sized, low-grade, ductal carcinoma in situ (DCIS) (Table 1A).

Table 1.

Diagnosis using definitive threshold of ΔFI (exploratory study).

Classification	Range of ΔFI	IC	NIC	Total cancer	Non-malignant	Total	Predictive value
							Positive	Negative
A. 5 min ΔFI
Positive (+)	≧ 0.101	21 (80.8%)	5 (62.5%)	26 (76.5%)	1(1.7%)	27 (29.0%)	96.3% (26/27)
Uncertain	0.00586 ~ 0.101	5 (19.2%)	2 (25.0%)	7 (20.6%)	41 (69.5%)	48 (51.6%)
Negative (-)	< 0.00586	0 (0%)	1 (12.5%)	1 (2.9%)	17 (28.8%)*	18 (19.4%)		94.4% (17/18)
Total		26 (100%)	8 (100%)	34 (100%)	59 (100%)	93 (100%)
B. 20 min ΔFI
Positive (+)	≧ 0.831	21 (80.8%)	5 (62.5%)	26 (76.5%)	1(1.7%)	27 (29.0%)	96.3% (26/27)
Uncertain	0.175 ~ 0.831	5 (19.2%)	2 (25.0%)	7 (20.6%)	17 (28.8%)	24 (25.8%)
Negative (-)	< 0.175	0 (0%)	1 (12.5%)	1 (2.9%)	41 (69.5%)*	42 (45.2%)		97.6% (41/42)
Total		26 (100%)	8 (100%)	34 (100%)	59 (100%)	93 (100%)

The number of samples in which ΔFI is in each range is indicated. Percentages of IC, NIC, Total cancer, and Non-malignant samples are indicated in parentheses. The definitive negative and positive thresholds were determined using data points with a 3% false negative rate and a 3% false positive rate.

IC, invasive cancer; NIC, non-invasive cancer; ΔFI, change in fluorescent intensity.

The negative diagnostic capacity at 20 min was significantly superior to that at 5 min (* p < 0.001).

Among the 27 samples categorized as positive (+), only 1 was non-malignant tissue, yielding a positive predictive value (PPV) of 96.3% (26/27) (Table 1A). Of the 18 cases categorized as negative (-), only 1 was cancer (NIC), resulting in a negative predictive value (NPV) of 94.4% (17/18) (Table 1A).

Comparison with cancer detection accuracy of GGT probe

We compared the predictive value function of the 5 min ΔFI to that of the previously reported GGT probe. As presented in Table 2, the definitive positive rate of the α-Man probe with 5 min ΔFI was significantly superior to that of the GGT probe (23.0% vs. 76.5%), whereas the definitive negative rate was almost equivalent (1.8% vs. 2.9%). These results indicated that the uncertainty rate was significantly reduced with the use of the α-Man probe (20.6%) compared to the GGT probe (75.2%; p < 0.0001). Moreover, the PPV and NPV of the α-Man probe were equivalent to those of the GGT probe (92.9% and 97.0%, respectively)^18,19.

Table 2.

Comparison of diagnostic values of α-Man and GGT probes using the definitive threshold of 5 min ΔFI (exploratory study).

5 min ΔFI		GGT	α-Man
Positive (+)	Definitive positive zone	26 (23.0%) *	26 (76.5%) *
Uncertain	Pending zone	85 (75.2%) *	7 (20.6%) *
Negative (-)	Definitive negative zone	2 (1.8%)	1 (2.9%)
Total number of cancers		113 (100%)	34 (100%)

The number of samples in each ΔFI range is shown, with the corresponding percentage of the total number of samples given in parentheses. Data for 113 cancer cases analyzed using the GGT probe in the multicenter study were obtained from our previous study^19,20. The definitive negative and positive thresholds were determined using data points with a 3% false negative rate and a 3% false positive rate. Significant differences were observed compared to those in GGT data, as determined using Pearson’s chi-square test.

IC, invasive cancer; NIC, non-invasive cancer; ΔFI, change in fluorescent intensity.

* p < 0.001.

These results showed that the cancer detection accuracy of the α-Man probe was superior to that of the GGT probe.

Determination of optimal measurement time to reduce the rate of uncertain cases

As the rate of uncertain cases at the 5 min ΔFI was high (51.6%), we investigated the optimal measurement time required to reduce this rate. At each measurement time (3, 5, 10, 15, 20, and 30 min), the threshold that best discriminated cancerous from non-malignant tissues was determined by maximizing the Youden index on the ROC curve. The area under the ROC curve (AUC) increased in a time-dependent manner and reached a plateau at 15–20 min, while the sensitivity reached a plateau at 20 min (Fig. 4). Therefore, 20 min was considered to be the optimal measurement time of ΔFI to reduce the uncertainty rate.

Fig. 4.

Determination of measurement time and threshold value. Measurement time and threshold value were determined based on the results of the first exploratory study. The area under the ROC curve (AUC), sensitivity, specificity, and accuracy rate are plotted against the measurement time for the ΔFI.

Analysis according to definitive threshold of 20 min ΔFI

Based on the definitive threshold of 20 min ΔFI, 93 samples were categorized as follows: 27 (29.0%) were positive (+), 24 (25.8%) were uncertain, and 42 (45.2%) were negative (-) (Table 1B). Among the 34 cancers, 26 (76.5%) were positive (+), 7 (20.6%) were uncertain, and 1 case (2.9%) was negative (-). The negative case was a small-sized, low-grade DCIS (Table 1B).

Of the 27 positive (+) cases, only 1 was non-malignant tissue, yielding a PPV of 96.3% (26/27), whereas among 42 negative (-) cases, only 1 case was cancer (NIC), resulting in an NPV of 97.6% (41/42) (Table 1B).

The results were similar to those obtained with the 5-min ΔFI, except there was an increase in the negative rate (19.4 to 45.2%) and a corresponding decrease in the uncertainty rate (51.6 to 25.8%) with high PPV and NPV.

Second multicenter study

The objective of this second multicenter study was to validate the diagnostic capability of the α-Man probe for detecting BC demonstrated in the initial study, through a multicenter study (Fig. 1).

Samples in the multicenter study

We evaluated 330 samples from four institutes (Supplementary Table S1). After excluding samples for which there was a disagreement about the pathological diagnosis from the central committee (Supplementary Table S1), 303 samples were prepared as the FAS. Furthermore, 56 samples with errors in measurement, protocol, and preparation were excluded, and 274 samples (118 malignant and 156 benign) were analyzed as the PPS. The following analysis was performed using the PPS.

Analysis according to definitive thresholds of 5 and 20 min ΔFI

In the multicenter study involving 274 cases (118 cancers), the application of the 5 min ΔFI threshold classified 66 cases (24.1%) as positive (+), 147 cases (53.6%) as uncertain, and 61 cases (22.3%) as negative (-) (Fig. 5A, Table 3A). Among 118 cancers, 62 (52.5%) were positive (+), 53 (44.9%) were uncertain, and 3 cases (2.5%) were negative (-) (Table 3A).

Fig. 5.

Quantitative differences in ΔFI among different lesion types and the definitive thresholds of ΔFI (multicenter study). (A) 5 min ΔFI in various lesions plotted against the pathological classification of each in invasive cancer (IC), non-invasive cancer (NIC), and benign lesions (Non-malignant). The ΔFI of IC and NIC was significantly higher than that of benign (p < 0.001, Wilcoxon rank-sum test). Solid (0.0234) and dashed (0.116) lines represent the definitive negative and positive thresholds, respectively, as determined using data points with a 3% false negative rate and a 3% false positive rate. (B) 20 min ΔFI in various lesions plotted against the pathological classification of each sample. Green and blue lines represent the definitive negative and positive thresholds, respectively, as determined using data points with a 3% false negative rate and a 3% false positive rate.

Table 3.

Diagnosis using definitive threshold of ΔFI (multicenter study).

Classification	Range of ΔFΙ	IC	NIC	Total cancer	Non-malignant	Total	Predictive value
							Positive	Negative
A. 5 min ΔFI
Positive (+)	≧ 0.116	58 (61.1%)	4 (17.4%)	62 (52.5%)	4 (2.6%)	66 (24.1%)	93.9% (62/66)
Uncertain	0.0234–0.116	37 (38.9%)	16 (69.6%)	53 (44.9%)	94 (60.3%)	147 (53.6%)
Negative (-)	< 0.0234	0 (0%)	3 (13.0%)	3 (2.5%)	58 (37.2%) *	61 (22.3%)		95.1% (58/61)
Total		95 (100%)	23 (100%)	118 (100%)	156 (100%)	274 (100%)
B. 20 min ΔFI
Positive (+)	≧ 0.629	65 (68.4%)	8 (34.8%)	73 (61.9%)	4 (2.6%)	77 (28.1%)	94.8% (73/77)
Uncertain	0.214–0.629	29 (30.5%)	13 (56.5%)	42 (35.6%)	63 (40.4%)	105 (38.3%)
Negative (-)	< 0.214	1 (1.1%)	2 (7.7%)	3 (2.5%)	89 (57.1%) *	92 (33.6%)		96.7% (89/92)
Total		95 (100%)	23 (100%)	118 (100%)	156 (100%)	274 (100%)

The number of samples in which ΔFI is in each range is indicated. Percentages of IC, NIC, Total cancer, and Non-malignant samples are indicated in parentheses. The definitive negative and positive thresholds were determined using data points with a 3% false negative rate and a 3% false positive rate.

The negative diagnostic capacity at 20 min was significantly superior to that at 5 min (* p < 0.001).

IC, invasive cancer; NIC, non-invasive cancer; ΔFI, change in fluorescent intensity.

Of the 66 positive (+) cases, 4 were non-malignant tissues, yielding a PPV of 93.9% (62/66). Of the 61 negative (-) cases, 3 were NIC, resulting in an NPV of 95.1% (58/61) (Table 3A).

Using the same threshold criteria for 20 min ΔFI, the 274 cases were categorized as follows: 77 (28.1%) positive (+), 105 (38.3%) uncertain, and 92 (33.6%) negative (-) (Fig. 5B, Table 3B). Among the 118 cancers, 73 (61.9%) were positive (+), 42 (35.6%) were uncertain, and 3 cases (2.5%) were negative (-) (Table 3B).

Among the 77 positive (+) cases, 4 were non-malignant tissues, providing a PPV of 94.8% (73/77). Of the 92 negative (-) cases, 3 were cancers (1 of small IC and 2 of low-grade NIC), yielding an NPV of 96.7% (89/92) (Table 3B).

In addition, we confirmed the reliability of the standardized protocol by comparing the 5 min and 20 min ΔFI of cancerous samples among the four institutes. No significant differences in ΔFI were detected among the four institutes (Supplementary Fig. S2).

The results of this multicenter study were consistent with those of the initial study, demonstrating a high diagnostic accuracy with the 5 min ΔFI and a reduction in the number of uncertain cases with the 20 min ΔFI.

Relationship between ΔFI and BC subtypes or clinicopathological features

We found no significant differences in the 5 min and 20 min ΔFI among the BC subtypes or clinicopathological features (Supplementary Table S2). These findings indicate that the fluorescent α-Man probe can be applied to all types of BC, except for mucinous carcinoma.

Discussion

Fluorescence-guided detection is a promising alternative to IFSA. It offers advantages such as simplicity, a lower cost, the need for less space, and rapid analysis, without requiring a pathologist. It also allows further pathological examination. Therefore, we aimed to evaluate the diagnostic performance of the α-Man probe for intraoperative surgical margin assessment during BCS in BC. The present study demonstrates that the α-Man fluorescent probe has the potential to enable accurate intraoperative identification of both positive and negative surgical margins during BCS in BC. Compared with our previously developed GGT probe, the α-Man probe provides clearer diagnostic outcomes, supporting its potential as an effective tool for fluorescence-guided surgery.

The α-Man probe demonstrated high diagnostic accuracy within just 5 min. When the 5-min ΔFI exceeded the definitive positive threshold, a high PPV was achieved, whereas values below the definitive negative threshold yielded a high NPV. Notably, all invasive cancer samples exceeded the definitive negative threshold, indicating that a 5-min ΔFI below this threshold reliably excludes the presence of IC at the surgical margin. This finding supports the potential use of the α-Man probe to exclude IFSA in cases where surgeons prioritize the exclusion of residual IC. Although a few false negative cases involving NIC were observed, these were limited to small, low-grade lesions for which postoperative radiation may suffice. Notably, this probe is applicable to any BC except for mucinous carcinoma. Collectively, the rapid and reliable performance of the α-Man probe suggests its great potential for clinical implementation as a real-time intraoperative tool for surgical margin assessment during BCS.

One limitation of the 5-min ΔFI assessment is the relatively high rate of diagnostically uncertain cases. To overcome this limitation, we propose a two-step diagnostic strategy in which uncertain cases at 5 min are re-evaluated at 20 min. In the present study, this approach increased the proportion of correctly identified benign samples and reduced the uncertainty rate, while maintaining high diagnostic accuracy. This improvement may help reduce unnecessary re-excisions by enabling more definitive intraoperative margin assessment. Based on this strategy, we developed a practical diagnostic flow (Fig. 6), enabling appropriate intraoperative decisions while avoiding unnecessary surgery. In this framework, a “positive” result suggests the need for immediate re-excision, whereas a “negative” finding supports the completion of surgery without further resection.

Fig. 6.

Flow chart of two-step fluorescent diagnosis at 5 and 20 min using the α-Man probe.

Furthermore, for future clinical application, the cut-off value from the second multicenter study will be validated in a prospective clinical trial using clinically relevant endpoints.

In addition to its diagnostic performance, the α-Man probe–based fluorescence method offers several practical advantages over conventional IFSA. This approach enables the rapid assessment of multiple margin samples simultaneously without requiring the presence of a pathologist, thereby streamlining the intraoperative workflow. Notably, the fluorescence procedure preserves tissue integrity for subsequent pathological evaluation. These features are particularly valuable given the growing shortage of pathologists, which is a relevant concern in Japan, where many cancer care institutions have limited or no pathology support^14,15. As the two-step diagnostic strategy reduces reliance on IFSA, this method may help alleviate the workload of pathologists and serve as a practical alternative for intraoperative margin assessment, especially in resource-constrained settings. Although it cannot fully replace pathological assessment, its implementation could reduce the number of cases requiring frozen section analysis, as additional pathological evaluation would mainly be needed for samples remaining “uncertain” at 20 min.

However, despite the two-step strategy, many cases remained uncertain even at 20 min, highlighting a key limitation of the current approach. Therefore, to improve diagnostic precision, combining the α-Man probe with additional fluorescent probes targeting different markers may be effective. We previously identified several such probes^23–25, and the assay system established in this study provides a practical foundation for future multi-probe applications. Furthermore, the biological role of MAN2C1 in cancers, including BC, remains largely unclear. Despite its implications in cancer progression and prostate carcinogenesis^27–29, further studies are needed to elucidate the functional relevance of MAN2C1 in BC.

In conclusion, the newly developed fluorescent probe, α-Man, may serve as a feasible tool for both negative and positive surgical margin assessment during BCS. This approach increases the number of sampling sites from the surgical margin, saves time, and reduces the burden on pathologists. For clinical implementation, a prospective clinical trial comparing intraoperative fluorescence findings with final pathological diagnoses will be essential to validate this approach, with key endpoints including diagnostic accuracy, false-negative rate, and re-excision rate.

Materials and methods

Clinical samples

All procedures were performed in accordance with the ethical standards of the Declaration of Helsinki, and the study protocols were approved by the Ethics Review Committees of Osaka Chiken Hospital (reference number: 008ERC, 009ERC), Shonan Memorial Hospital (22 − 01), JCHO Kurume General Hospital (No. 250), Showa University Koto Toyosu Hospital (22-057-A), National Defense Medical College (4672), Kawasaki Medical School (2022 − 0195), Kurume University Medical Center (22098), and Mie University Hospital (H2022-142). This multicenter study was registered with the Japan Registry of Clinical Trials (jRCT number: jRCT1051240034; registration date: May 16, 2024). All participants provided written informed consent prior to sampling. Inclusion criteria were as follows: (1) diagnosed with breast cancer and scheduled for surgery, and (2) provided informed consent.

Resected tissue samples included malignant lesions as well as benign and non-malignant breast tissues obtained from the surgical specimens.

Axillary lymph node dissection and sentinel lymph node biopsy (SLNB) were performed when clinically indicated. The SLNB method (dye or radiocolloid) was determined by each participating institution. The dyes used in this study (Indigo carmine or Patent Blue V) do not emit fluorescence, and radiocolloids (e.g., 99mTc) are detected by gamma radiation. Therefore, these agents do not interfere with the fluorescence measurements of the α-mannosidase probe. As a precaution, areas with residual dye were excluded from the analysis.

Tissue samples were excluded based on predefined technical and pathological criteria. Samples containing mucinous carcinoma or skeletal muscle were excluded because mucinous components may interfere with probe activation. Samples showing discordant morphology between visible-light images and corresponding HE sections were also excluded. For pathological review, four board-certified pathologists independently evaluated all specimens. When one pathologist provided an indeterminate judgment, the final diagnosis was determined from the remaining three. When one diagnosis differed from the other three, the sample was considered pathologically discordant and excluded. In addition, samples with errors in fluorescence measurement, protocol deviations, or inadequate preparation were excluded from the per-protocol set. Details are provided in the Supplementary Table S1.

The first exploratory study enrolled 46 participants who underwent mastectomy or BCS between April and September 2021 at the Ueo Breast Cancer Hospital. The second multicenter study enrolled 126 patients who underwent surgery between May and October 2022 at the Ueo Breast Cancer Hospital and three other hospitals.

Sample preparation and fluorescence detection

Using the method described in our previous report^19,20 (Supplementary Fig. S3), samples were obtained from three sites of the resected specimen: the central portion of the BC, its peripheral lesion, which comprises non-invasive cancer, and the distal portion of normal mammary tissue. The α-mannosidase–targeted fluorescent probe (α-Man probe) was provided by Goryo Chemical, Inc. (Sapporo, Japan). The probe has an excitation wavelength of 460 nm and an emission wavelength of 525 nm. The trademark and patent for this probe have been granted.

After sampling from the surgical margin, the specimen was carefully immersed in the measurement well with the evaluated surface facing upward. The fluorescence intensity of each sample was then compared with the corresponding histopathological findings. Each sample (at most 22 × 22 × 6 mm) was first moistened with saline and then soaked in a phosphate-buffered solution for 1–4 min to remove blood, fat, and other extracellular materials. Each sample was then placed in a well (35 mm in diameter), to which 2 mL of the probe solution was added. Fluorescence imaging was initiated immediately after the addition of the probe solution. During imaging, the samples were soaked in the probe solution.

Fluorescence images were recorded using a dedicated apparatus (Hamamatsu Photonics, Hamamatsu, Japan) with 460 nm excitation lights, a 525 nm emission filter, a built-in camera, a software program, and eight sample wells. Color images of the samples were recorded under visible light. Changes in fluorescence intensity (ΔFI) were calculated by subtracting the fluorescence intensity (FI) measured just after the probe administration. The ΔFI at 30 min was recorded in each sample (Supplementary Fig. S4). The total time required for analysis was within 15 min for the 5-minute protocol and within 30 min for the 20-minute protocol.

Detailed information about the excluded samples, including the number of cases, reasons for exclusion, associated pathological findings, and distribution across participating centers, is provided in the Supplementary Table S1.

Pathological examination

Formalin-fixed, paraffin-embedded (FFPE) tissue specimens were prepared from the same samples used for fluorescence measurements. Immediately after conducting the fluorescence assessment, each specimen was fixed using standard formalin fixation and subsequently processed for routine H&E staining. Histopathological sections were prepared from the same surface that had been subjected to fluorescence measurement. The morphological features observed in the visible light images obtained during fluorescence imaging were carefully compared to those of the pathological specimens. Samples exhibiting discrepancies in morphological features were excluded from the analysis.

A pathologist examined the HE-stained specimens in the first exploratory study performed at the Ueo Breast Cancer Hospital. In the second multicenter study conducted across four hospitals, pairs of visible and HE-stained images were randomly numbered and provided to four pathologists in the central committee without any background information, and they independently examined all specimens. Each sample was classified into four groups: invasive cancer (IC), non-invasive cancer (NIC), benign proliferative lesions, and normal breast tissue. IC included invasive ductal carcinoma and invasive lobular carcinoma; NIC included ductal carcinoma in situ and lobular carcinoma in situ; benign proliferative lesions included precancerous lesions (flat epithelial atypia, atypical lobular hyperplasia, and atypical ductal hyperplasia), benign tumors (fibroadenoma), and adenosis group (adenosis, ductal and lobular hyperplasia, and columnar cell lesions); and normal breast tissue included mammary glands and fat. The pathological diagnosis was considered consistent when at least three pathologists classified the sample into the same category. Samples with inconsistent pathological diagnoses were excluded from the analysis.

Statistical analysis

Statistical analyses were performed using Microsoft Excel 2019 and JMP 13 (SAS Institute Inc., Cary, NC, USA). The Wilcoxon rank-sum test (for two groups) or Kruskal–Wallis test (for more than two groups) was used to compare the mean ΔFI values between cancer and benign lesions and examine the relationship between ΔFI and clinicopathological features. Pearson’s chi-square test was used to compare the diagnostic value of the definitive thresholds of 5 min ΔFI between the α-Man and GGT probes. Statistical significance was set at p < 0.05. The receiver operating characteristic (ROC) curve analysis was used to set the respective thresholds.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

Hiroki Ueo, Hiroaki Ueo, Shun Kunii, and Yui Nagamori performed the experiments and analyses. Hiroki Ueo, Hiroaki Ueo, and Takaaki Masuda wrote the manuscript. Hiroki Ueo, Hiroaki Ueo, Takako Doi, Miki Yamaguchi, Sadako Akashi-Tanaka, Tomoko Takamaru, Yoko Kubota, and Yuichiro Ka approved the protocol, performed the breast surgery, and collected the clinical materials. Ayako Gamachi, Hitoshi Tsuda, Takuya Moriya, Rin Yamaguchi, Yuji Kozuka, and Takeshi Sasaki determined the pathological procedures and examined the samples. Koshi Mimori, Yuki Ando, Kyohhei Fujita, Yasuteru Urano, and Takaaki Masuda supervised the study. All authors read and approved the final manuscript.

Funding

This study was financially supported by Goryo Chemical, Inc. Hamamatsu Photonics K.K. provided the fluorescence measurement instruments.

Data availability

All data generated or analyzed during this study are included in this published article and its supplementary material files.

Declarations

Competing interests

Shun Kunii and Yui Nagamori are employees of Goryo Chemical, Inc. The study was financially supported by Goryo Chemical, Inc.Hamamatsu Photonics K.K. provided the fluorescence measurement instruments.The remaining authors declare no competing interests.